This competitive renewal proposal extends and enriches our longitudinal investigation of suicidal behavior in borderline personality disorder (BPD), focusing specifically on medically serious suicidal acts. Our work to date points to the central role of a socio-emotional diathesis to suicide in BPD. In this cohort of patients with BPD ascertained over the previous 15+ years (and other related studies of suicide), we have found that (i) negative affectivity and aggression robustly predict suicidal behavior, (ii) that socio-emotional influences disrupt fronto-limbic brain activity subserving cognitive control in BPD, and (iii) that suicidal behavior is associated with behavioral and neurocomputational alterations in social and non-social decision-making. The proposed research builds on clinical theory and experimental data that characterize BPD patients as interpersonally hypersensitive, often distorting social cues, forming extreme opinions of others, and misattributing malevolence. Perceived rejection, abandonment, or betrayal trigger aggression and negative emotional responses that can lead to a suicidal crisis. To explicate further the mechanisms underlying this escalation, we will test a conceptual model that views maladaptive responses to social stressors in terms of Pavlovian interference with goal-directed decision processes. We will thus test the overarching hypothesis that, in an interpersonal context, emotionally potent cues overshadow patients? representations of strategic goals and distort the expected value of their actions. This work bridges three units of analysis: prospective predictors of high-lethality suicide attempts in our high-risk cohort, the focus of much of our prior research (Aim 1; between- persons); behavioral and affective signatures of the socio-emotional diathesis to suicidal behavior in daily life captured via experience sampling (Aim 2; within-person); and neural underpinnings of maladaptive social decision-making assessed by functional MRI augmented with a reinforcement learning computational model (Aim 3; within-person). A key strength of continuing work with this BPD cohort with a longitudinally ascertained high risk of suicidal behavior is that many participants are now entering midlife, when impulsive suicidal acts become less common, but the risk of serious suicidal acts increases. This is also a time of transition for our longitudinal study with Aims 2 and 3 introducing significant innovation into the work and with new investigators contributing expertise in suicide risk (Dombrovski, Szanto), BPD and experience sampling (Hallquist, Pilkonis, Wright), imaging and computational modeling (Dombrovski, Hallquist), and quantitative methods (Hallquist, Wright). The proposed study addresses the #1 question of the Prioritized Research Agenda for Suicide Prevention, Why do people become suicidal? Neurocomputational constructs of prediction error (in a social context) and expected value assessed in Aim 3 directly reflect the framework of the Research Domain Criteria (RDoC), as do our conceptual and analytic efforts to bridge units of analysis. This work will open new avenues for discovery of behavioral, affective, and neural markers elucidating processes proximal to the suicidal act.